Abstract MP22: Age-Varying Implications of Recalibration of Cardiovascular Disease Risk Prediction in a New England-Based Healthcare System

Abstract

Introduction: Atherosclerotic cardiovascular disease (ASCVD) risk estimation via the 2013 Pooled Cohort Equations (PCE) overestimates risk in population-based cohorts. We evaluated its recalibration performance in a contemporary healthcare cohort. Methods: We identified 160,926 patients in the Mass General Brigham healthcare system aged 40-79 years and without prior ASCVD or lipid-lowering medication use between 2010 and 2012. Recalibration was achieved using sex- and race-specific baseline survival and risk factor distributions of the derivation (80%) set. Internal validation compared Kaplan-Meier-adjusted observed and predicted risk based on original versus recalibrated PCE. Changes in sensitivity and specificity were calculated across various treatment thresholds for each age decile. Results: Of the 160,926 patients (mean [SD] age: 54.6 [8.6] years; 61.4% female; 82.7% non-Hispanic White), 20,373 (12.7%) underwent ASCVD over 10 years. The original PCE underestimated ASCVD risk (observed vs. predicted incidence rate: 0.13 vs. 0.05) globally and across all sex, race, and risk categories. After recalibration (observed vs. predicted incidence rate: 0.13 vs. 0.18), the prediction closely mirrored the actual ASCVD incidence in low-to-borderline risk but overestimated risk in the high risk. At ≥7.5% treatment threshold, patients aged 40-49 years derived the greatest net relative benefit (sensitivity: +42.9%; specificity: -26.6%) whereas the extent of specificity loss surpassed that of sensitivity gain beyond age 50 years with minimal reclassification by seventh decade in life. Treatment threshold yielding the greatest net benefit was incrementally higher with older age from ≥7.4% in age 40-49 years to ≥24.1% in 60-69 years. Conclusion: In healthcare setting, PCE significantly underestimated risk. Recalibration recovered sensitivity at the expense of reduced specificity in age-differential manner, highlighting the need for individualized prediction and clinical decision-making.